Evaluating and tuning system response in the MFTF-B control and diagnostics computers

The software system running on the Supervisory Control and Diagnostics System (SCDS) of MFTF-B is, for the major part, an event driven one. Regular, periodic polling of sensors' outputs takes place only at the local level, in the sensors' corresponding local control microcomputers (LCC's). An LCC reports a sensor's value to the supervisory computer only if there was a significant change. This report is passed as a message, routed among and acted upon by a network of applications and systems tasks within the supervisory computer (SCDS). Commands from the operator's console are similarly routed through a network of tasks, but in the opposite direction to the experiment's hardware. In a network such as this, response time is partially determined by system traffic. Because the hardware of MFTF-B will not be connected to the computer system for another two years, the authors are using the local control computers to simulate the event driven traffic that we expect to see during MFTF-B operation. In this paper they show how we are using the simulator to measure and evaluate response, loading, throughput, and utilization of components within the computer system. Measurement of the system under simulation allows us to identify bottlenecks and verify their unloosening. They also use the traffic simulators to evaluate prototypes of different algorithms for selected tasks, comparing their responses under the spectrum of traffic intensities